Data analysts and other computer users often interact with and request data from computer-based databases containing large collections of data objects. The data object collections may be generated and collected by various corporations, businesses, governmental agencies, and other organizations and often represent huge amounts of data. For example, the data may include DNS traffic logs, transaction logs of banks, call data records, or any other potentially high-volume data.
One particular type of data that such organizations may often collect and analyze is geospatial data. Geospatial data generally represents any data that includes a geographical component and that associates the data with particular geographic locations or areas. For example, a crime enforcement agency may track data that corresponds to crime reports received within a particular city. One item of information stored in association with each crime report may include a geographical component indicating a location of the incident specified by the crime report. An example data object representing each crime report may include values representing a longitude and latitude corresponding to the geographical location of the incident. A data analyst with access to such data may then specify geospatial search requests such as, for example, a search request for crime reports associated with incidents within a specified geographical radius or within a four block perimeter.
Organizations also frequently collect data that includes a temporal component. A temporal component generally refers to one or more data values that indicate a particular point in time, such as a date and/or time of day, or a range of time. For example, in reference to the example crime report information, a crime enforcement agency may store in association with each crime report a time value indicating an approximate time the incident in the crime report occurred. In the example, by storing the temporal information, data analysts may specify requests for data objects associated with a particular time range of interest, for example, all crime reports associated with an incident that occurred within the past three months.
Databases have provided mechanisms for efficiently indexing and searching geospatial data for a number of years. These mechanisms include geospatial indices that transform the geospatial data into values that are well-suited for database indexing while preserving a notion of geographical locality in the indexed data. However, indexing and searching for geo-temporal data presents a number of challenges that are not resolved by today's geospatial indexing techniques. For example, because the notion of geographical locality does not accurately translate to the added dimensionality of time, the use of geospatial indexing mechanisms on geo-temporal data often results in highly inefficient processing of geo-temporal queries. Further, user requests for geo-temporal data are often focused on highly varied granularities of time for reasons practical to desired lines of inquiries on the data, but that complicate efforts to efficiently index the data.
The present disclosure attempts to address these problems and others, facilitating low latency searches of large and dynamic data sets that include geo-temporal data.
The approaches described in this section are approaches that could be pursued, but not necessarily approaches that have been previously conceived or pursued. Therefore, unless otherwise indicated, it should not be assumed that any of the approaches described in this section qualify as prior art merely by virtue of their inclusion in this section.